Backend Engineer & Applied AI

Orchestrating Agentic AI & high-scale backends.

I engineer high-throughput microservices handling billions of requests at Airbnb and build robust, production-grade agentic workflows with RAG and multi-agent frameworks.

Ayush Ranjan

About Me

I’m Ayush Ranjan, a backend engineer based in San Francisco. I build practical AI and backend systems: robust digital bridges that connect products with millions of users.

I earned my B.Tech in IT from Manipal University Jaipur and completed my master's degree in Computer Science in 2025 at the University of California, Santa Cruz (UCSC), specializing in AI systems and database architecture.

Currently, I'm a Backend Developer at Altimetrik, engineering core billing flows and microservices on the Airbnb Payments Wallet & Instruments team (and previously on the Incentives team). Previously, I led projects at Capgemini for Mercedes-Benz (winning 3rd place at Innocircle 2022) and researched RAG architectures and wearable AI agents in UCSC's academic labs.

At UCSC, I also served as a Teaching Assistant for Database Systems across four terms and Software Engineering for one term, mentoring students in SQL optimization, database design, and practical engineering workflows.

Beyond systems design, I'm a passionate football player, an avid reader of engineering publications, and a cook who treats kitchen experimentation like optimizing distributed pipelines.

ayush@system:~ - bash
Welcome to Ayush's Interactive Terminal.
Type '--help' to see available query options.

ayush@system:~ $

Work Experience

Backend Developer

Altimetrik (Client: Airbnb Payments)
Sep 2025 - Present
  • Engineered high-concurrency Java microservices processing 2.55B monthly requests (avg 5,060 QPS) and architected wallet true-name collection pipelines for fraud analytics, CVV check flows, and background cleanup jobs via Kafka and Tempo.
Java Dropwizard Kafka Thrift RPC MySQL Redis Tempo Kubernetes Bazel

Graduate Researcher

AI Explainability & Accountability (AIEA) Lab, UCSC
Oct 2024 - Aug 2025
  • Researched advanced cyclic RAG workflows (Corrective, Adaptive) using RAGAS scoring to achieve a 35-50% improvement across evaluation metrics in document-retrieval pipelines deployed via Docker, Kubernetes, and FastAPI.
Python LangGraph RAG FastAPI Docker Kubernetes

AI Research Intern

IRKM Lab, UCSC (Stealth Hardware Startup)
Jul 2024 - Sep 2024
  • Built a 0-to-1 multimodal AI agent for smart audio-visual wearables using Dialogflow, LangGraph, and Pinecone to execute semantic intent routing with 95% classification accuracy.
Python LangGraph Whisper Pinecone Flask Dialogflow

Associate Consultant & Senior Analyst

Capgemini (Clients: Mercedes-Benz & Arek Oy)
Jul 2021 - Aug 2023
  • Managed schema evolution for Mercedes-Benz's vehicle diagnostic system (decreasing configuration import latency by 66.67%), built Autosar mock testing telemetry systems, and won 3rd place at the Innocircle Innovation hackathon.
Java Spring Boot React Redux Micro-Frontends IBM Db2 MySQL Autosar

Senior Analyst Intern

Capgemini Technology Services
Jan 2021 - May 2021
  • Engineered an enterprise online medical portal as a Java Full Stack Developer, creating secure React interfaces and Spring Boot REST controllers integrated via Axios with extensive JUnit testing.
Java Spring Boot React Axios JUnit MySQL

Academic Background

M.S. in Computer Science
GPA: 3.92 / 4.00
University of California, Santa Cruz
2023 - 2025
Core Subject Focus
  • Distributed Systems & Concurrency
  • Multi-Agent AI Workflows & LangGraph Orchestration
  • Retrieval-Augmented Generation (RAG) Architecture
  • Advanced Machine Learning & Deep Learning
  • Database Systems Design & Query Optimization
B.Tech. in Information Technology
GPA: 8.03 / 10.00
Manipal University Jaipur
2017 - 2021
Core Subject Focus
  • Data Structures & Algorithms
  • Operating Systems & Concurrency
  • Computer Networks & Security
  • Database Management Systems (DBMS)
  • Object-Oriented Software Design (Java / C++)

Key Projects

Trimmr: LLM Token Saver

Trimmr: LLM Token Optimizer

Built a Rust CLI proxy that intercepts verbose git command output and compacts it before it reaches an LLM context window - reducing token consumption by up to 74%. Filters git status, diff, log, and more while preserving exit codes for safe automation.

Rust CLI LLM Token Optimization
HTMLDiff: DOM-Aware HTML Diff Viewer

HTMLDiff: DOM-Aware Diff Viewer

Built a local DOM-aware diff tool for HTML files that visualizes actual rendered changes - not raw code diffs. Features split-view with scroll sync, pure CSS overlays (no DOM injection), a 10-snapshot time-travel history ring, J/K keyboard navigation, and a Chrome MV3 extension for one-keystroke workflow integration.

Node.js Chrome Extension DOM Diffing MV3
Fathom: Code-Aware Search

Fathom: Code-Aware Search

Designed a code-aware search engine combining semantic (Tree-sitter), structural (SCIP), and literal queries behind FastAPI for LLMs. Created SQLite Librarian and ChromaDB indices with MCP-compatible access for coding agents.

FastAPI ChromaDB Tree-sitter MCP
CLIP Glitches

CLIP Image Encoding Bugs

Analyzed image encoding errors in CLIP using the Discrepancy Analysis Framework (DAF) and DINOv2. Discovered 14 systemic faults, including four novel issues. Received an A+ in Neural Computation at UCSC.

Python DINOv2 CLIP PyTorch
Video to MP3

Video-to-MP3 Microservices

Built a containerized pipeline with a JWT authentication gateway, RabbitMQ queues, MongoDB GridFS binary storage, and converter workers orchestrated with Kubernetes in Minikube.

RabbitMQ MongoDB Kubernetes Docker
Sentiment CNN

CNN Sentiment Classifier

Implemented a multi-filter CNN (sizes 2, 3, 4, 5) in PyTorch to capture n-gram patterns. Tokenized with spaCy and initialized with GloVe embeddings, achieving 87% test accuracy.

PyTorch SpaCy CNN GloVe
Image Captioning

Attention Image Captioner

Built encoder-decoder LSTM framework with ResNet50 vision extraction. Integrated additive and dot-product attention modules evaluated on validation loss and BLEU metrics.

PyTorch ResNet50 Attention LSTM
Facial Attendance

Facial Attendance System

Built an OpenCV-based face tracking and recognition workflow with a custom Tkinter administration dashboard and Google Text-to-Speech feedback.

OpenCV Tkinter Python

Programming Stack

Hover over any tech to highlight related work experience and projects.

Backend Architecture

Java Spring Boot Dropwizard Thrift RPC Kafka Microservices ETL Pipelines XML Processing Maven / Gradle

AI & Data Layer

Python LangGraph RAG Systems Pinecone ChromaDB PyTorch MySQL IBM Db2 Redis Whisper AI

DevOps & Tools

Kubernetes Docker FastAPI CI/CD Pipelines Bazel React Redux MCP Tools Autosar

Awards & Achievements

2023
UCSC Kaggle Competition Winner
2022
3rd Place at Innocircle 2022, Mercedes-Benz Internal Innovation Hackathon
2022
AZ-900 Microsoft Azure Fundamentals Certification (Score: 940/1000)
2021
Java Skill Certification by HackerRank
2014
Gold Medal Winner (School Topper), 16th National Science Olympiad (SOF)

Favorite Books

Code by Charles Petzold
How Google Works by Eric Schmidt & J. Rosenberg
The Facebook Effect by David Kirkpatrick
Shiva Trilogy by Amish Tripathi

My GitHub Activity

Get In Touch

I am open to discussions about scalable backend systems, agentic AI and RAG, research collaborations, or career opportunities. Drop me a line!

Location

San Francisco, CA, USA

He patrols. He turns. He jumps for fish.